ADVOSC

ADVOSC is a desktop app for VRChat OSC. It started because I wanted a nicer chatbox than the ones I could find, and it grew from there: a chatbox template engine you can build with blocks or write by hand, avatar parameter control and profiles, avatar scaling, and an OSC forwarder.

Windows only, since a few parts (media info, trackers, hardware sensors) talk to Windows APIs directly.

Grab a build from Releases, or come say hi on Discord if something breaks.

Run ADVOSC-Setup.exe and it installs and opens itself, there is no wizard to click through. After that it checks for updates on its own and downloads them in the background, then asks whether you want to restart. If you would rather not have that, use the zip instead: it is the same app, it just never updates itself.

---

The chatbox

Everything in the chatbox is a template. You write text, and anywhere you want live data you drop in a placeholder like {{MediaInfo;Track}}. ADVOSC re-renders it every 2.2 seconds (VRChat's rate limit) and sends the result.

There are two ways to write one, and they both end up as the same template underneath, so you can start in one and finish in the other.

Simple editor

You stack blocks, reorder them, and watch the preview update as you go. No syntax to memorize. There are 30 block types:

Category	Blocks
Basic	Text, Text Transform, Text Replace, Text Truncate, Text Pad, Text Fallback, Current Time, New Line
Media	Now Playing, Song Progress, Heart Rate
Display	Progress Bar, Health Bar, Star Rating, Toggle Text, Number Format
Animate	Marquee / Scroll, Bounce, Typewriter, Blink, Cycle Texts
Logic	Condition
VRChat	Hotkey State, Avatar Param, VR Tracker
Tools	Stopwatch, Session Time, Shortcut, Number Calc, Random Number

The part that makes it actually usable for real templates: most blocks accept more than a plain value as their source. You can feed them:

• a literal like 42, true, or some text
• an OSC address like /avatar/parameters/Health
• an inner placeholder like [[MediaInfo:Duration]]
• a full placeholder like {{Shortcut;Time}}

So mixing media data, tracker battery, heart rate, hotkeys and conditions in one visual flow works fine. A few things people tend to build:

• a now playing line that falls back to something else when nothing is playing
• a song progress bar with a moving head character
• a status line driven by a hotkey or an avatar parameter
• a weather line with a condition emoji, temperature and sunset time
• long media titles truncated so they fit, HP padded so the layout stops jumping
• numbers clamped, mapped into another range, rounded, or randomized
• marquee, blink, typewriter, rotating messages
• AFK or stream status that swaps text depending on a condition

Advanced editor

If you would rather just write the template, this is a real editor with autocomplete for every module and placeholder. Nothing is hidden from you here, and anything the simple editor can produce you can also write by hand.

Whichever editor you have open is the one that gets sent.

My Templates

A saved template holds the whole setup: the template text, both editors' state, and every module's settings. So switching templates swaps your entire chatbox in one click instead of you rebuilding it.

Action	What it does
Save Current As Template	Snapshots your current setup under a name
Use	Switches to a saved template, and can save your current one first so you can come back
Overwrite With Current Setup	Updates a saved template in place after you tweak something
Share	Gives you a compact share code, the raw JSON, or a .advosc.json file
Import	Takes a share code, template JSON, template file, or an old "Export All Settings" bundle

Sharing strips your credentials. Heart rate tokens, API keys, widget ids, custom WebSocket URLs and shortcuts you marked hidden never leave your machine, and the share window tells you exactly what it removed. There is a checkbox to include them on purpose if you are exporting a backup for yourself.

Imported templates just land in your list. Nothing switches until you say so.

---

Modules

Both editors run on the same 15 modules. Click one to see what it does.

🎵 Media Info :: whatever you are listening to

Reads Windows' own media session, so it works with Spotify, browsers, most players. Synced lyrics included when they exist.

Placeholder	Output	Description
{{MediaInfo;Status}}	Playing	Playback status (Playing, Paused, Stopped)
{{MediaInfo;Track}}	Never Gonna Give You Up	Track title
{{MediaInfo;Artist}}	Rick Astley	Artist name
{{MediaInfo;Album}}	Whenever You Need Somebody	Album name
{{MediaInfo;Position}}	120000	Current position in milliseconds
{{MediaInfo;Duration}}	300000	Total duration in milliseconds
{{MediaInfo;AppName}}	Spotify	Which app is playing it
{{MediaInfo;Lyric}}	Never gonna give you up	Current synced lyric line

🕒 Time :: clocks, dates, durations

Placeholder	Output	Description
{{Time;NowMillis}}	1700000000000	Current time in milliseconds (Unix epoch)
{{Time;Now;HH:mm}}	14:30	Current time, your format
{{Time;Now;yyyy-MM-dd}}	2023-11-14	Current date, your format
{{Time;Timezone}}	America/New_York	System timezone
{{Time;UTCOffset}}	+00:00	Current UTC offset
{{Time;FormatDuration;3600000;Short}}	1h 0m 0s	Milliseconds as a readable duration
{{Time;ElapsedMillis;...}}	5000	Time since a timestamp

🔤 Text :: transforms, bars, animations

The biggest module. Transform text, build meters out of it, animate it.

Placeholder	Output	Description
{{Text;Upper;hello}}	HELLO	Uppercase
{{Text;Lower;HELLO}}	hello	Lowercase
{{Text;Title;hello world}}	Hello World	Title case
{{Text;Trim; hello }}	hello	Strip surrounding whitespace
{{Text;Replace;cat;dog;cat nap}}	dog nap	Replace every match
{{Text;Length;hello}}	5	Length
{{Text;Reverse;hello}}	olleh	Reverse it
{{Text;Repeat;3;Hi }}	Hi Hi Hi	Repeat N times
{{Text;Slice;0;5;Hello World}}	Hello	Substring
{{Text;Format;Rounded;text}}	ⓣⓔⓧⓣ	Unicode styling
{{Text;Format;Bold;text}}	𝐭𝐞𝐱𝐭	One of the newer styles
{{Text;Truncate;10;Long text...}}	Long text...	Cut with an ellipsis
{{Text;Build;ProgressBar;30;100;10;█;░;▓}}	██▓░░░░░░░	Progress bar, optional head character
{{Text;Build;HealthBar;3;5;♥;♡}}	♥♥♥♡♡	Hearts style meter
{{Text;Build;StarRating;3;5;★;☆}}	★★★☆☆	Rating display
{{Text;Build;Toggle;true;ON;OFF}}	ON	Truthy or falsy to text
{{Text;NumberFormat;2;.;,;1234.5}}	1,234.50	Format a number for display
{{Text;Animate;Marquee;...}}	scrolling	Marquee
{{Text;Animate;Typewriter;Hello}}	H → He → Hel	Reveal over time
{{Text;Animate;Blink;ON;OFF}}	ON → OFF	Alternate two texts
{{Text;Animate;EachOne;A;B;C}}	A → B → C	Cycle through items

🔢 Number :: math and randomness

Placeholder	Output	Description
{{Number;Random;Int;1;100}}	42	Random integer in a range
{{Number;Random;Float;0;1}}	0.7531	Random float in a range
{{Number;Clamp;150;0;100}}	100	Keep a value inside bounds
{{Number;Map;5;0;10;0;100}}	50	Remap one range onto another
{{Number;Floor;3.7}}	3	Round down
{{Number;Ceil;3.2}}	4	Round up
{{Number;Round;3.5}}	4	Round to nearest
{{Number;Abs;-5}}	5	Absolute value

🧪 Expression :: conditions and real math

When a condition gets too weird for the other modules, write it out.

Placeholder	Output	Description
{{Expr;5 > 3;Yes;No}}	Yes	If / else
{{Expr;Math.sqrt(16)}}	4	Math expressions
{{Expr;Math.sin(Math.PI/2)}}	1	Trig
{{Expr;[[MediaInfo:Status]]=='Playing';🎵;⏸️}}	🎵	Conditions on live data

❤️ Heart Rate :: Pulsoid, HypeRate, Stromno, custom

Add your feeds under Modules → Heart Rate. Each one gets a name that you use in the placeholder, so your token never ends up sitting inside a template you might share.

Platform	What you need
Bluetooth (BLE)	A band or strap broadcasting heart rate. Hit Scan and pick it from the list
Pulsoid	Access token
HypeRate	API key + session id
Stromno	Widget id
Custom WebSocket	WebSocket URL + optional JSON path (like data.heartRate)

Placeholder	Output	Description
{{HeartRate;SOURCE;HeartRate}}	75	Current heart rate
{{HeartRate;SOURCE;IsOnline}}	true	Connection status
{{HeartRate;SOURCE;AverageHR;300}}	72	Average over N seconds (max 900)
{{HeartRate;SOURCE;MaxHR}}	120	Session maximum
{{HeartRate;SOURCE;MinHR}}	55	Session minimum
{{HeartRate;SOURCE;Battery}}	84	Device battery percentage (Bluetooth only)
{{HeartRate;SOURCE;HasContact}}	true	Skin contact (Bluetooth only)
{{HeartRate;SOURCE;DeviceName}}	Xiaomi Smart Band 10 Pro	Advertised device name (Bluetooth only)
{{HeartRate;SOURCE;SensorLocation}}	Wrist	Where the device is worn (Bluetooth only)

Bluetooth notes. Anything implementing the standard heart rate service (0x180D) works: Coospo, Polar, Garmin, Amazfit, Xiaomi and so on. There is no pairing step, and you should not pair the device in Windows Settings, since a bonded device that is not broadcasting is a common cause of stuck connections.

Your device has to actually be broadcasting. On Xiaomi bands that is Settings → Heart rate → Bluetooth broadcast, and it is off by default. While broadcasting, the band drops its connection to your phone, so if the phone app grabs it first the PC will never see it. Broadcast usually stops when the band screen goes back to the watchface. Drops are normal and ADVOSC reconnects on its own with a backoff.

Old templates using {{Pulsoid;TOKEN;…}} migrate themselves: the token becomes a named source and the template gets rewritten. Pulsoid still works as an alias too.

📡 OSC Data :: raw avatar parameters

Placeholder	Output	Description
{{OSCData;/avatar/parameters/AFK}}	true	Any OSC parameter value
{{OSCData;/avatar/parameters/VelocityX}}	0.5	Floats too

🎮 OpenVR Trackers :: battery and status

Placeholder	Output	Description
{{OVRTrackers;BatteryLevel;finder}}	85	Battery level (0-100)
{{OVRTrackers;IsCharging;finder}}	true	Charging or not
{{OVRTrackers;ModelNumber;finder}}	Vive Tracker 3.0	Model
{{OVRTrackers;SerialNumber;finder}}	LHR-12345678	Serial
{{OVRTrackers;IsExists;finder}}	true	Whether it is there at all

The finder can be a device index, serial number, device class, or model number.

⚙️ Process :: is it running, and for how long

Placeholder	Output	Description
{{Process;IsRunning;VRChat.exe}}	true	Running or not
{{Process;StartedAt;VRChat.exe}}	1700000000000	Start timestamp
{{Process;SessionTime;VRChat.exe}}	3600000	Time since it started

🖥️ System Resources :: CPU, GPU, memory, network

What your PC is doing right now, sampled by a native helper twice a second.

CPU

Placeholder	Output	Description
{{System;CPUName}}	AMD Ryzen 9 7950X 16-Core Processor	Processor name
{{System;CPUUsage}}	13.7	Overall usage percentage
{{System;CPUCoreUsage;0}}	24.5	One logical core, zero based index
{{System;CPUCores}}	16	Physical core count
{{System;CPUThreads}}	32	Thread count
{{System;CPUFrequency}}	4501	Current frequency in MHz
{{System;CPUTemperature}}	62	Celsius, empty when Windows exposes no sensor

Memory

Placeholder	Output	Description
{{System;MemoryUsed}}	31.2 GB	In use
{{System;MemoryTotal}}	127.2 GB	Total physical memory
{{System;MemoryAvailable}}	95.9 GB	Available to applications
{{System;MemoryUsage}}	24.6	Usage percentage
{{System;SwapUsed}}	0 B	Page file in use
{{System;SwapTotal}}	8 GB	Total page file size
{{System;SwapUsage}}	0.0	Page file usage percentage

GPU. Every GPU placeholder takes an optional finder as its first parameter: a zero based index, or part of the name or vendor. Leave it empty and you get the GPU with the most VRAM, which is the discrete card on a laptop or APU system.

Placeholder	Output	Description
{{System;GPUName}}	NVIDIA GeForce RTX 3070 Ti	GPU name
{{System;GPUVendor}}	NVIDIA	NVIDIA, AMD, Intel or Unknown
{{System;GPUUsage}}	42.0	Utilization percentage
{{System;VRAMUsed}}	2.7 GB	VRAM in use
{{System;VRAMTotal}}	8 GB	Total VRAM
{{System;VRAMUsage}}	34.3	VRAM usage percentage
{{System;GPUTemperature}}	53	Celsius, NVIDIA only
{{System;GPUPower}}	52.1	Watts, NVIDIA only
{{System;GPUFanSpeed}}	0	Fan percentage, NVIDIA only
{{System;GPUCoreClock}}	1020	Core clock in MHz, NVIDIA only
{{System;GPUCount}}	2	How many GPUs were found
{{System;GPUUsage;intel}}	3.0	Usage of whichever GPU matches the finder

Network and uptime

Placeholder	Output	Description
{{System;Upload}}	1.5 MB/s	Current upload speed
{{System;Download}}	12.1 MB/s	Current download speed
{{System;TotalUploaded}}	32.6 GB	Uploaded since boot
{{System;TotalDownloaded}}	24.0 GB	Downloaded since boot
{{System;Uptime}}	22h 41m	Since boot, also takes seconds, minutes, hours, days or clock

Units and decimals. Every byte and speed placeholder takes a unit and a decimal count. With auto (the default) the value scales itself and keeps its label. Give it an explicit unit and you get a bare number, which is what you want when feeding it into math or a progress bar.

Placeholder	Output	Description
{{System;MemoryUsed}}	31.2 GB	Auto unit, one decimal
{{System;MemoryUsed;GB}}	31.2	Fixed unit, number only
{{System;MemoryUsed;GB;0}}	31	Fixed unit and decimal count
{{System;CPUUsage;0}}	14	Percentages take a decimal count directly
{{System;Download;MB;2}}	12.15	MB/s as a bare number

Heads up: GPU temperature, power, fan speed and core clock come from NVIDIA's NVML, so they are empty on AMD and Intel. Usage and VRAM work everywhere. CPU temperature needs a sensor Windows actually exposes, which it usually does not.

🌤️ Weather :: current conditions and forecast

Live weather and a 7 day forecast from Open-Meteo, free and no API key needed. Save your locations under Modules → Weather. Each gets a name, and the first one becomes the default that gets used when you leave the location parameter empty.

The location parameter takes a saved name, any place name (it gets looked up for you), or lat,lon coordinates.

Right now

Placeholder	Output	Description
{{Weather;Temperature}}	21.4	Temperature at your default location
{{Weather;Temperature;Home}}	21.4	Temperature at a saved location
{{Weather;FeelsLike;Tokyo}}	19.8	Apparent temperature
{{Weather;Condition}}	Partly cloudy	Condition as text
{{Weather;Emoji}}	⛅	Condition as an emoji, day and night aware
{{Weather;Code}}	2	WMO weather code
{{Weather;Humidity}}	63	Relative humidity percent
{{Weather;Precipitation}}	0.2	Precipitation of the last hour
{{Weather;Rain}}	0.2	Rain and showers of the last hour
{{Weather;Snowfall}}	0	Snowfall of the last hour
{{Weather;WindSpeed}}	12	Wind speed
{{Weather;WindGusts}}	24	Gusts
{{Weather;WindDirection}}	270	Direction in degrees
{{Weather;WindCompass}}	W	Direction as a compass point
{{Weather;CloudCover}}	48	Cloud cover percent
{{Weather;Pressure}}	1013	Surface pressure in hPa
{{Weather;IsDay}}	true	Daytime at that location

Forecast. The last parameter is the day offset: 0 today, 1 tomorrow, up to 6.

Placeholder	Output	Description
{{Weather;High;;0}}	24.1	Day's high
{{Weather;Low;;0}}	12.6	Day's low
{{Weather;Sunrise;;0}}	1700000000000	Sunrise timestamp in milliseconds
{{Weather;Sunset;;0}}	1700000000000	Sunset timestamp in milliseconds
{{Weather;UVIndex;;0}}	5.3	Max UV index of the day
{{Weather;PrecipitationChance;;1}}	40	Chance of precipitation percent
{{Weather;PrecipitationSum;;1}}	3.4	Total precipitation for the day
{{Weather;DailyCondition;;1}}	Slight rain	Condition text for the whole day
{{Weather;DailyEmoji;;1}}	🌦️	Condition emoji for the whole day
{{Weather;Date;;1}}	2023-11-15	Local calendar date

Location and status

Placeholder	Output	Description
{{Weather;Location}}	Berlin, Berlin, DE	Resolved location name
{{Weather;Timezone}}	Europe/Berlin	Timezone there
{{Weather;UpdatedAt}}	1700000000000	When it was last fetched
{{Weather;IsOnline}}	true	Whether the last fetch worked
{{Weather;Unit;Temperature}}	°C	Unit symbol for Temperature, Wind or Precipitation

Units (Celsius / Fahrenheit, km/h / m/s / mph / knots, mm / inch) and the refresh interval live in the module tab. Timestamps pair up with the Time module:

{{Weather;Emoji}} {{Weather;Temperature}}{{Weather;Unit;Temperature}} • sunset {{Time;Timestamp;[[Weather:Sunset]];HH:mm}}

🌐 HTTP & WebSocket :: pull in anything else

If a module for it does not exist, make one out of this. Add endpoints under Modules → HTTP & WebSocket and each gets a name you use in placeholders.

Two kinds of source:

• HTTP Request. Any method with your own headers and body. Polled on the interval you pick, and only while a template is actually using it.
• WebSocket. Stays connected and reconnects on its own. Placeholders read the latest message that arrived. You can send a subscribe or auth message on connect, and a keepalive on a timer.

The URL, headers and body can all contain placeholders, filled in right before the request goes out, so a request can be built from live data.

Reading values. Bodies get parsed as JSON and the path parameter walks into them: data.items[0].title. Array indexes can be negative to count from the end ([-1] is the last), quoted keys handle weird names (["odd key"]), and length works on arrays. Empty path gives you the whole body. Anything that is not JSON comes back as plain text.

Placeholder	Output	Description
{{Request;Value;MyApi;data.value}}	42	Value at a JSON path in the latest body
{{Request;Value;MyApi;items[0].name}}	Sword	Array indexing, negatives count from the end
{{Request;Value;MyApi}}	{"data":{"value":42}}	Whole body when the path is empty
{{Request;Text;MyApi}}	{"data":{"value":42}}	Raw body exactly as received
{{Request;Has;MyApi;data.value}}	true	Whether that path exists
{{Request;Get;https://api.example.com/x;data.value}}	42	One-off GET, no source to configure
{{Request;Get;https://api.example.com/x;data.value;300}}	42	Same, with a refresh interval in seconds

Status

Placeholder	Output	Description
{{Request;IsOnline;MyApi}}	true	Last request worked, or the socket is connected
{{Request;Status;MyApi}}	200	HTTP status code, empty for WebSocket sources
{{Request;Ok;MyApi}}	true	Whether that status was 2xx
{{Request;Error;MyApi}}	Request timed out after 10000ms	Last error, empty when things are fine
{{Request;Count;MyApi}}	7	Responses received, or messages that arrived
{{Request;UpdatedAt;MyApi}}	1700000000000	When data last arrived, in milliseconds
{{Request;Age;MyApi}}	12	Seconds since data last arrived
{{Request;Header;MyApi;content-type}}	application/json	A header from the latest response

Pair it with Expr to show something sensible while a source is down:

{{Expr;[[Request:IsOnline:MyApi]]=='true';Players online: [[Request:Value:MyApi:server.players]];Server offline}}

Requests are made by ADVOSC itself and not by the page, so browser CORS rules do not apply and any endpoint works. Bodies are capped at 2 MB. A source is only polled while a template or the module tab is using it, and a source that is already failing never blocks your chatbox while it retries.

⌨️ Hotkey :: press a key, change the chatbox

Placeholder	Output	Description
{{Hotkey;IsPressed;MyHotkey;1000}}	true	Pressed within a timeout
{{Hotkey;IsToggled;MyHotkey}}	false	Toggle state, press to flip

⏱️ Stopwatch :: timers driven by hotkeys

Stopwatches you control with hotkeys from the Hotkey module. Good for AFK detection and session tracking.

Placeholder	Output	Description
{{Stopwatch;ElapsedMs;MyTimer}}	125000	Elapsed milliseconds
{{Stopwatch;IsRunning;MyTimer}}	true	Currently running
{{Stopwatch;IsPaused;MyTimer}}	false	Currently paused

The hotkey actions are Start/Toggle, Pause (keeps the time), Reset (back to zero, keeps running), Stop (reset and pause), and Reset + Start, which is the one you want for AFK timers.

📝 Shortcut :: name your own placeholders

Give a name to a chunk of placeholder soup you keep retyping, then use the name instead. The simple editor also generates these for you behind the scenes.

🎤 Speech :: talk, and put it in the chatbox

Speech to text, with optional live translation. Electron has no speech recognition of its own, so ADVOSC starts a small server on your own machine and Chrome does the listening. Open the page from the Speech module tab, allow the microphone, and keep the tab open. Nothing leaves your machine except what Chrome sends to its own recognition service, and whatever the translation provider needs.

Placeholder	Output	Description
{{Speech;Text}}	hey how are you doing	Everything heard right now, including the words still being spoken
{{Speech;Final}}	hey how are you doing	Only the finished sentences
{{Speech;Interim}}	and then i said	Only the part still being spoken
{{Speech;Last}}	how are you doing	The most recent finished sentence
{{Speech;Translated}}	hey nasılsın	Translation of what is currently heard
{{Speech;TranslatedFinal}}	hey nasılsın	Translation of the finished sentences only
{{Speech;Both; | }}	hey how are you doing | hey nasılsın	Original and translation, with your separator
{{Speech;IsListening}}	true	The browser page has the mic open
{{Speech;IsSpeaking}}	true	Words are coming in right now
{{Speech;HasText}}	true	There is something to show
{{Speech;IsMuted}}	false	VRChat reports your own mic as muted
{{Speech;Age}}	3200	Milliseconds since the last thing you said
{{Speech;Confidence}}	0.92	Recognizer confidence for the last sentence, 0 to 1
{{Speech;Language}}	en-US	Recognition language
{{Speech;TargetLanguage}}	tr	Language it translates into
{{Speech;DetectedLanguage}}	en	What the translation provider detected
{{Speech;Provider}}	Google Translate	Translation provider in use
{{Speech;IsServerRunning}}	true	The local speech server is up
{{Speech;IsPageConnected}}	true	The browser page is open and connected
{{Speech;ServerUrl}}	http://127.0.0.1:7274/?t=...	Address of the recognition page
{{Speech;Error}}		Last recognition or translation error, empty when fine

Translation supports four providers. Google works with no API key at all, so you can turn it on and it just goes. DeepL usually sounds the most natural. OpenRouter and Gemini run the text through an LLM, which handles slang and context better, and you can steer them with your own extra instructions. Keys are stored on your machine and are left out of template exports.

Text clears itself after a while so the chatbox does not keep showing something you said ten minutes ago. You can change the delay, how many sentences to keep at once, and a character cap, all on the module tab.

There is also an option to follow VRChat's own mute. Turn on "Clear and stop capturing while muted in VRChat" and muting yourself in game wipes the current text and closes the browser mic, then unmuting reopens it. It reads VRChat's MuteSelf parameter, so it needs OSC enabled in game like everything else here.

A line that only appears while you are talking:

{{Expr;[[Speech:HasText]]=='true';🎤 [[Speech:Both; | ]]}}

🔊 Sound :: play sound effects from your library

Plays sounds you added in the Sound Effects tab. The action placeholders render as nothing, so you can drop them anywhere in a template. They fire once when the optional condition turns true, then wait for it to go false again. Without a condition they fire once when the placeholder shows up. A placeholder that stays in your template does not replay every 2.2 seconds.

Placeholder	Output	Description
{{Sound;Play;boop;condition}}		Plays boop when the condition turns true
{{Sound;Loop;boop;condition}}		Loops boop while the condition is true, stops it when it turns false
{{Sound;Stop;boop;condition}}		Stops every playing copy of boop
{{Sound;StopAll;condition}}		Stops every sound
{{Sound;Volume;50;condition}}		Sets the master volume, 0 to 100
{{Sound;IsPlaying;boop}}	true	boop is playing right now
{{Sound;Playing}}	boop, vine boom	Names of everything playing right now

Play a boop sound whenever someone boops you:

{{Sound;Play;boop;[[OSCData:/avatar/parameters/Booped]]}}

New to the syntax? The Placeholder Learning Guide walks through the two-layer system from scratch, and the expert one goes further once that clicks.

---

Beyond the chatbox

Avatar OSC control

Every parameter on your avatar, in one list. You can lock a toggle so nothing changes it by accident, route one parameter's value into another, or animate one for breathing effects and color shifts.

Avatar profiles

Save a whole parameter state (outfit, toggles, props, whatever mood you had going) and load it back in one click. Profiles export and import as JSON, and you can search or filter by avatar once you have collected a few too many.

Avatar scale

Resize yourself in game by sending an eye height to VRChat, anywhere from 0.01 to 10.0 meters. Drag the slider or type an exact number, save your own named presets, and optionally have ADVOSC remember a height per avatar so it comes back on avatar change or instance join.

There is also parameter forwarding, so you can drive your size from inside your avatar: /avatar/parameters/advosc_eyeheight goes to /avatar/eyeheight, either normalized 0..1 onto a range you pick, or straight through as meters.

OSC Forwarder

Take any chatbox template value and send it to an OSC address on a timer. The value field accepts any placeholder, so {{OSCData;/avatar/parameters/X}} or {{Time;Now;HH:mm}} both work. Pick the target path by hand or straight from the current avatar's schema, cast to Float, Int, Bool or String, optionally remap numbers from one range to another (0..1 to 0..255, say), and give each rule its own interval down to 100 ms.

Sound Effects

A soundboard that reacts to your avatar. It lives in its own tab in the chatbox editor, next to the OSC Forwarder.

Add audio files to the library and give each one a name. Each sound gets a play button, its own volume, a cooldown so it can't be spammed, an optional global hotkey (press Esc on the hotkey button to clear it), and an option to send an OSC value while it plays (for example SoundPlaying=true, then false when it ends) so an avatar animation can follow along. Sounds layer on top of each other if you trigger them again. The files stay where they are on disk and are not part of template exports.

Triggers are the fun part. A trigger has a template, a match, and a list of sounds. When the template result starts matching, one of the sounds plays at random. By default it plays to the end, but a trigger can also loop the sound while it matches, or cut it off the moment it stops matching. The match can be a plain value (1, true, idle) or start with an operator (> 0.5, != 0, <= 3). Triggers are checked the moment OSC comes in, so short parameter flips still get caught.

You can pick the output device, so a virtual audio cable can send the sounds into your VRChat mic. There's also a master volume, a mute button, a Stop all button with its own hotkey, and an option to stop everything and hold new sounds while you're muted in VRChat.

---

Coming soon

Local speech recognition with Whisper, as an offline alternative to the browser recognizer in the Speech module.

---

Development

Bun, Electron, Svelte 5, and a bit of Rust for the Windows-only bits.

bun install     # deps
bun run dev     # main + preload + renderer + electron, all watching
bun run build   # production build into dist/
bun run package # distributable

---

GPL-3.0. Made for the VRChat community, mostly at hours I should have been asleep.